Build a quantum-safe backbone for AI with F5 and NetApp

Steve GormanSolutions Architect

AI model training that uses S3 to aggregate tremendous amounts of source content doesn’t just need massive numbers of GPUs; it needs a backbone that features high availability, load balancing, high-performance data throughput, and storage control to maximize GPU utilization and eliminate expensive bottlenecks. Data storage also needs advanced security that can keep pace with AI’s speed and complexity while future-proofing sensitive datasets with post-quantum cryptography (PQC).

With F5 Application Delivery and Security Platform (ADSP) and NetApp StorageGRID, enterprises get fast, scalable, and PQC-ready solutions to handle the demanding workloads of data ingestion for AI model training while preparing for the quantum future.

F5 enhances NetApp StorageGRID with efficiency and resiliency

The GPU bandwidth needs for AI workloads can reach hundreds of gigabytes per second, and training datasets can take up petabytes of storage. F5 BIG-IP, a core component of F5 ADSP that offers a range of traffic management and security capabilities, provides the hyperscale needed to intelligently distribute those content-transfer workloads, ensuring GPUs aren’t sitting idle while waiting for bandwidth or storage availability.

F5 BIG-IP Local Traffic Manager (LTM) is a workhorse with all the tools to provide the throughput and security to load-balance layer 4 and layer 7 traffic with features such as FastL4, which uses hardware acceleration through embedded Pack Velocity Acceleration (ePVA) chips, or the OneConnect feature to reduce excessive TCP connections. These features can be configured with profiles and further tuned with customized layer 7 functions using F5 iRules.

Meanwhile, companies with extremely large datasets have realized that legacy network attached storage (NAS) solutions using traditional file-sharing protocols like SMB and NFS are unable to meet the scale required for AI workloads and are moving to S3 on HTTPS. NetApp StorageGRID provides S3 object storage that can be used for AI data, apps, and services at scale, allowing near-instant access to buckets, both writing and copying objects that can number in the billions. BIG-IP, which can be deployed as a highly available solution, manages NetApp clusters by monitoring and removing unhealthy nodes with failed HTTP checks and directing writes to other healthy node members.

When paired together, BIG-IP and NetApp StorageGRID equip enterprises with an advanced architecture that is fast, resilient, secure, and ready to meet the challenges of AI workloads.

Prepare for Q-Day with PQC in F5 and NetApp storage clusters

All of the devices you use today, such as your laptop and smartphone as well as the servers that host and secure your company data, operate on binary-based processors that encode data in ones and zeroes. These are known as classical computers which depend on classical encryption methods that we use to secure our data today.

Quantum computing doesn’t store data in ones and zeroes. It uses quantum bits, better known as qubits, which can handle significantly larger and faster amounts of data. As quantum computing develops, it’s predicted that it will eventually compromise and fully break our current encryption standards by 2034, which the industry refers to as Q-Day.

While 2034 may feel like a long time from now, hackers are stealing data today and waiting until they can decrypt it in the future. These types of breaches are commonly referred to as “harvest now, decrypt later” attacks. Companies can start preparing for these attacks now by implementing TLS 1.3 to support PQC in their technology stacks. While the PQC algorithms themselves are compute-intensive and should only be used on your most sensitive and high-risk data, these algorithms also need to be planned and tested with key stakeholders in your organization.

Until the computational overhead of PQC is reduced, classical encryption will be ideal for most workloads. Companies need to slowly introduce PQC for their most sensitive and long-lived data such as tax and legal documents and health records. This level of control and testing requires the platform intelligence and granularity that BIG-IP delivers. When a request is made to S3 storage clusters, BIG-IP can utilize F5 iRules to determine the data classification and risk, apply the appropriate encryption, and route the traffic to proper NetApp StorageGRID nodes.

Deploy fast, future-ready AI model training with F5 and NetApp

By implementing these solutions, organizations can reap the full benefits of their AI deployments while helping to secure themselves from current, hidden threats posed by attackers harvesting their data now to exploit later when it can be decrypted. Together, F5 and NetApp provide the platform capabilities needed to simplify the path to fast, efficient, and future-ready AI.

To learn more, visit our F5 and NetApp solutions webpage. Also, be sure to read our press release.

About the Author

Steve GormanSolutions Architect

More blogs by Steve Gorman